CLAYTON@XRT.UPENN.EDU ("Clayton, Paul D.") (08/21/87)
Information From TSO Financial - The Saga Continues... Chapter 16 - August 9, 1987 There was a question from the Netherlands concerning the use of system common dump file, SYSDUMP.DMP, and the impact of a multi system shutdown/crash. The following error was received. $ ANALAYZE/CRASH_DUMP SYS$SYSTEM:SYSDUMP.DMP %SDA-E-NOREAD, unable to access location 80002C58 -SDA-E-NOTVALID, information not in physical memory The dump file in the SYS$COMMON area was said to be the size of the largest memory configuration, plus 4 (I hope). There is a very vital sequence of events that occur when a dump file is USED and I believe is the problem here. The writing of memory contents to the dump file is done as a 'last gasp' type of operation. The 'normal' device driver is NOT used to perform the function and thereby all cooridination with other systems for access to this file is NOT done. The result is that if multiple systems shutdown/crashed at the same time the net effect can be a 'mixture' of the different systems. This would happen on systems that have more memory then can be written in one I/O to the disk. If the memory is small enough to write with one I/O, then you would get the last system to perform the write in the dump file. This is the primary reason I have seperate dump files for the systems that are critical. By eliminating the over writing, some knowledge can be gained about the cause of the crash. Paul D. Clayton - Manager Of Systems TSO Financial - Horsham, Pa. USA Address - CLAYTON%XRT@CIS.UPENN.EDU